Past attempts have been made to predict ice cover characteristics and behavior along rivers, both during at freeze-up and breakup conditions; however these attempts perform with varying success and are for the most part site specific. This research introduces a geospatial modelling approach which can fulfil the task of improving the predictive power of river freeze-up and ice cover breakup events. A geospatial model clusters discretized sections of a river that have similar geomorphological (e.g. sinuosity, slope, width, etc.) and ice cover (e.g. ice thickness and type) features under certain hydraulic and meteorological conditions using a multi-variant statistical clustering technique such as the principle component analysis. These clusters can then be grouped into Geomorphic Response Units, each representing a particular set of ice cover characteristics and behavior along the river. RADARSAT-2 imagery and field sampling complement the work to help in the development of the geospatial model. The Slave River in Canada is used as a test site.